Social welfare and measurement of segregation (Q1308808)

This paper looks at measures of segregation and their economic interpretation. Measures of segregation are, among others, ``the exposure index'', ``the index of dissimilarity'' and ``the index of segregation'', the latter measuring in the case of occupational segregation by race, for example, the sum of the differences between the percentages of the black respective white labour force that are employed in a particular occupation. What are the properties that these measures of segregation have in common? The angle from which the problem is being considered is that of a planner who is to select an allocation. The planner is characterized by his preferences over allocations. The axioms that characterize a large class of measures of segregation are: (1) the planner's preferences are asymmetric and negatively transitive, (2) distribution equivalence (i.e. if two allocations induce the same distribution on quality pairs -- social composition and intrinsic qualities, let's say, then indifference must hold; (3) similarity irrelevance, a kind of independence condition; (4) distribution relevance. The first theorem in the paper states that under these four assumptions there exists a numerical representation which evaluates an allocation by the weighted sum of fraction of individuals who experience the social and material qualities of the categories. The second result of the paper provides conditions under which measures of segregation are consistent with standard economic criteria such as the welfaristic criterion of Pareto optimality. The second result shows that any measure of segregation consistent with the four axioms above satisfies Pareto optimality if and only if all individuals within a social class have identical preferences. This means that these measures reflect Pareto improvements if and only if there is no heterogeneity within a social class. Allowing for preference diversity within classes invalidates the result.